Interpretable Convolutional Neural Networks with Dual Local and Global Attention for Review Rating Prediction

By Sungyong Seo, Jing Huang, Hao Yang, Yan Liu. 2017

457 citations recorded by the source corpus.

Recently, many e-commerce websites have encouraged their users to rate shopping items and write review texts. This review information has been very useful for understanding user preferences and item properties, as well as enhancing the capability to make personalized recommendations of these websites. In this paper, we propose to model user preferences and item properties using convolutional neural networks (CNNs) with dual local and global attention, motivated by the superiority of CNNs to extract complex features. By using aggregated review texts from a user and aggregated review text for an item, our model can learn the unique features (embedding) of each user and each item. These features are then used to predict ratings. We train these user and item networks jointly which enable the interaction between users and items in a similar way as matrix factorization. The local attention provides us insight on a user's preferences or an item's properties. The global attention helps CNNs focus on the semantic meaning of the whole review text. Thus, the combined local and global attentions enable an interpretable and better-learned representation of users and items. We validate the proposed models by testing on popular review datasets in Yelp and Amazon and compare the results with matrix factorization (MF), the hidden factor and topical (HFT) model, and the recently proposed convolutional matrix factorization (ConvMF+). Our proposed CNNs with dual attention model outperforms HFT and ConvMF+ in terms of mean square errors (MSE). In addition, we compare the user/item embeddings learned from these models for classification and recommendation. These results also confirm the superior quality of user/item embeddings learned from our model.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Information Systems, Model Architectures

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Xiangnan He
2. Yongfeng Zhang
3. Xu Chen
4. Xiang Wang
5. Xing Xie
6. Tat-seng Chua
7. Meng Wang
8. Fei Sun

In Computer Vision and Pattern Recognition

1. Yongfeng Zhang
2. Xiangnan He
3. Xiang Wang
4. Xu Chen
5. Meng Wang
6. Yi Tay
7. Anh Tuan Luu
8. Siu Cheung Hui

In Information Systems

1. Xiangnan He
2. Yongfeng Zhang
3. Xu Chen
4. Xiang Wang
5. Xing Xie
6. Tat-seng Chua
7. Meng Wang
8. Le Wu

In Model Architectures

1. Aminu Da’u
2. Fei Sun
3. Tarek R. Besold
4. Xing Xie
5. Siu Cheung Hui
6. Yi Tay
7. Bin Cui
8. Roberto Confalonieri

Citations in the corpus

• Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI (2019, neutral): “Informativeness [5, 44, 21, 25, 26, 45, 30, 32, 34, 35, 37, 38, 41, 46, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 63, 64, 65, 66, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 86, 87, 88, 89, 59, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 1”
• RippleNet: Propagating User Preferences on the Knowledge Graph for Recommender Systems (2018, background): “The idea was soon transplanted to recommender systems [4, 23, 33, 36, 47].”
• Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI) (2018, neutral): “Examples include [102]–[104].”
• Explainable Recommendation: A Survey and New Perspectives (2018, background): “(b) The model generates attention scores for each review and highlight the words with high attention scores as explanations (Seo et al. [2017]).”
• A Context-Aware User-Item Representation Learning for Item Recommendation (2017, background): “is also in line with the experimental results reported in [10], Third, CARL consistently achieves the best MSE scores across the five datasets.”
• Techniques for interpretable machine learning (2018, methodology): “igned according to totally dierent principles and have various forms of implementations. Take the recommender system eld as an example, both interpretable latent topic models and attention mechanism [32] could provide some extent of interpretability. In this case, however, how can we compare the int”
• Multi-Pointer Co-Attention Networks for Recommendation (2018, methodology): “Recent advances in deep learning has spurred on various innovative models that exploit reviews for recommendation [2, 24, 37].”
• A review on deep learning for recommender systems: challenges and remedies (2018, unknown): “…Wang et al. (2016) Xu et al. (2017a) Volkovs et al. (2017) Rassweiler Filho et al. (2017) Vall et al. (2017) Nguyen et al. (2017) Hassan (2017) Seo et al. (2017b) Nedelec et al. (2017) Peska and Trojanova (2017) Deldjoo et al. (2017) Zhang et al. (2017b) Kim et al. (2016) model group…”
• Personalized Fashion Recommendation with Visual Explanations based on Multimodal Attention Network: Towards Visually Explainable Recommendation (2019, background): “By discovering such fine-grained user preference, our recommended items can also be explained visually, which is more vivid, convenient, attractive, and comprehensively-efficient as compared with traditional textual explanations [7, 26, 31, 46] for on line users in the fashion domain.”
• A Survey on Accuracy-Oriented Neural Recommendation: From Collaborative Filtering to Information-Rich Recommendation (2021, background): “ent pieces of content, such that informative elements are automatically selected for item content representation [91], [92], [93], [94], [95], [96], [97], [98].”
• In2Rec: Influence-based Interpretable Recommendation (2019, background): “Recently, due to the powerful ability in representation learning, deep learning is adopted in recommendation to model textual content and generate the explanations [7–9 , 29].”
• Coevolutionary Recommendation Model: Mutual Learning between Ratings and Reviews (2018, methodology): “Similar ideas have also proven helpful for extracting textual features from customer reviews; see for example [35].”
• Incorporating Interpretability into Latent Factor Models via Fast Influence Analysis (2018, unknown): “Existing works on enhancing recommender systems with interpretability can be classified into different categories based on the type of recommendation models, including latent factor models [1, 3, 23, 26, 39], memory-based models [29, 32], graph-based models [5, 15, 35], deep learning models [8, 24, ”
• A Survey on Deep Learning for Human Mobility (2020, background): “, machine translation [39, 185] and speech recognition tasks [41]), the usage of attention mechanisms rapidly extended to computer vision, healthcare, and recommendation systems [40, 165, 211].”
• Visually Explainable Recommendation (2018, background): “Existing explainable recommendation mod-elsusuallyinterprettherecommendationsbased onuserreviews[20, 25, 37].”
• A General Survey on Attention Mechanisms in Deep Learning (2022, background): “Similarly, in [61], both global and local alignment are used”
• Graph Neural Networks in Recommender Systems: A Survey (2020, background): “Due to the significance of explainability, many interests have focused on designing explainable recommendation models or conducting post-hoc interpretations [62, 125, 138].”